Re: Subplot demo performance

[Konstantinos Poulios]

On Wed, Feb 16, 2011 at 2:04 PM, Ben Abbott <address@hidden> wrote:
> On Feb 16, 2011, at 4:55 AM, Konstantinos Poulios wrote:
>
>> After recent changes in the calculation of axes
>> position/outerposition, Ben observed very long rendering delay for the
>> 1st subplot demo. Since I cannot reproduce this problem I start this
>> new thread in order to ask for some testing from others, so that we
>> can narrow down the cause for this regression.
>>
>> @Ben, since you are the one who can reproduce this issue I would
>> suggest to start with some more basic tests than the subplot demo, so
>> e.g. you can try:
>>
>> 1. Drawing a grid of 3x3 plot with no data and axes labels/titles.
>>
>> 2. Add some data to the plots
>>
>> 3. Add titles
>>
>> 4. Add axes labels
>>
>> Depending on the delay you observe in these tests we will be able to
>> narrow down the problem a bit.
>>
>> BR
>>
>> Kostas
>>
>> P.S. Once we resolve this I will open a new thread for the subplot layout 
>> issue.
>
> I've attached a script. The results for gnuplot are
>
> 1. Drawing a grid of 3x3 plot: 0.49 secs
> 2. Add some data to the plots: 1.78 secs
> 3. Add titles: 1.90 secs
> 4. Add axes labels: 5.21 secs
>
> For fltk, the results are nearly identical.
>
> 1. Drawing a grid of 3x3 plot: 0.50 secs
> 2. Add some data to the plots: 1.81 secs
> 3. Add titles: 1.93 secs
> 4. Add axes labels: 5.36 secs
>
> Ben
>
>

nice script, so most time is spent when labels are added. It is
strange however that on my system it runs significantly faster, and
there is no significant delay in comparison to octave 3.4.0. These are
my results with 3.4.0:

1. Drawing a grid of 3x3 plot: 0.09 secs
2. Add some data to the plots: 0.27 secs
3. Add titles: 0.29 secs
4. Add axes labels: 0.66 secs


And with the current development version:

1. Drawing a grid of 3x3 plot: 0.09 secs
2. Add some data to the plots: 0.26 secs
3. Add titles: 0.29 secs
4. Add axes labels: 0.73 secs

The results with gnuplot and fltk are identical. Extra delay is
observed only for test #4 and it is about 10% in comparison to 3.4.0.
Could you maybe provide results with 3.4.0 also? In any case five
seconds are too long and we should try to do some optimization.

Kostas